Camelia, the Perl 6 bug

IRC log for #darcs, 2011-04-20

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:13 Hyper-Core joined #darcs
00:16 roconnor joined #darcs
00:19 Hyper-Core left #darcs
00:51 iago left #darcs
00:54 JaffaCake left #darcs
01:18 jeltsch1 joined #darcs
01:19 jeltsch left #darcs
01:40 lispy left #darcs
01:59 secorp joined #darcs
02:38 owst http://hpaste.org/45862 makes me sad! This is happening with a malformed boring file...
02:39 owst boringRegexps :: IO [Regex], but the result is used in a pure function. I'd like to handle the malformed regex when it's read, if possible, but since the error happens at match-attempt time, I'd have to (say) match against an empty string, which seems nasty.
02:40 owst I can't see any useful 'tryParseRegex' functions - does anyone have any ideas?
02:53 intripoon_ joined #darcs
02:56 intripoon left #darcs
03:26 owst http://hpaste.org/45864/ is my (not working) attempt at a condensed version of the solution that I want to try (test against empty string). This still fails with the same error, can anyone spot the problem?
03:27 owst And on that note, bed!
03:28 gwern left #darcs
03:28 copumpkin joined #darcs
03:30 owst Err, stylisticly tweaked it a bit: http://hpaste.org/45865/
03:31 gwern joined #darcs
03:33 secorp left #darcs
03:35 owst left #darcs
03:59 secorp joined #darcs
04:03 exlevan left #darcs
04:10 exlevan joined #darcs
04:35 sm left #darcs
04:57 mornfall left #darcs
04:57 mornfall joined #darcs
04:57 mornfall left #darcs
04:57 mornfall joined #darcs
05:02 copumpkin left #darcs
06:23 lispy joined #darcs
07:05 shenshei joined #darcs
07:21 roconnor left #darcs
07:32 lelit joined #darcs
07:46 jeltsch1 left #darcs
07:52 bsrk2 joined #darcs
07:53 secorp left #darcs
07:56 jderque joined #darcs
08:13 jeltsch joined #darcs
08:15 Weltraumschaf joined #darcs
08:18 bsrk2 left #darcs
08:26 lispy left #darcs
08:35 jeffz left #darcs
09:08 balor joined #darcs
09:23 balor left #darcs
09:24 balor joined #darcs
09:25 kowey joined #darcs
09:30 kowey fossil's built-in ticket tracker in action: http://www.fossil-scm.org/index.html/rptview?rn=1
09:30 kowey using the name-tickets-w-hashes approaches
09:40 JaffaCake joined #darcs
09:45 kowey left #darcs
09:55 owst joined #darcs
10:13 npouillard left #darcs
10:18 npouillard joined #darcs
10:44 roconnor joined #darcs
11:16 owst left #darcs
12:35 jderque left #darcs
13:35 Gwern-away joined #darcs
13:35 Gwern-away left #darcs
13:35 Gwern-away joined #darcs
13:36 gwern left #darcs
13:46 balor left #darcs
13:47 Gwern-away is now known as gwern
13:47 balor joined #darcs
13:47 balor left #darcs
13:52 kowey joined #darcs
14:01 balor joined #darcs
14:11 raichoo joined #darcs
14:24 copumpkin joined #darcs
14:27 balor left #darcs
14:28 balor joined #darcs
14:59 arjanb joined #darcs
15:00 roconnor left #darcs
15:02 sm joined #darcs
15:12 balor left #darcs
15:33 raichoo left #darcs
15:56 lispy joined #darcs
16:10 jeltsch left #darcs
16:18 shenshei left #darcs
16:40 owst joined #darcs
17:07 owst left #darcs
17:13 Weltraumschaf left #darcs
17:22 mornfall (Oh, btw, if anyone noticed, I know the darcs buildbot is offline and it'll probably stay that way until we receive replacement drives, since both main system disks (RAID1) in the machine died.
17:22 mornfall )
17:23 owst joined #darcs
17:24 mornfall (The VM images have been live-mirrored to an external RAID5 which is OK, so restoring the service shouldn't be much work.)
17:24 mornfall We still need the disks though.
17:29 secorp joined #darcs
17:50 preflex left #darcs
17:51 copumpkin left #darcs
18:03 copumpkin joined #darcs
18:29 lispy left #darcs
18:30 dankna left #darcs
18:31 dankna joined #darcs
18:46 lispy joined #darcs
18:54 roconnor joined #darcs
18:55 roconnor is there any "fuzziness" in darcs' hunk matching, or is it always precise (like I expect)?
19:02 Igloo It's precise
19:03 * roconnor shakes his head at the fuzzy-matching-is-what-you-want people
19:03 teratorn applying a diff with `patch' it probably is, but yeh
19:04 roconnor teratorn: sure. I understand the case where you have no choice but to fuzzy match because you don't have the information needed.
19:05 Riastradh roconnor, the fuzzy-matching-is-what-you-want people probably shake their heads at you, too.  `Silly Darcs user.  Insists on some useless notion of precise matching that causes exponential time to merge.'
19:06 roconnor Riastradh: oh right, the exponential time thing.
19:06 roconnor That's not a totally unreasonable reply
19:07 roconnor though I think that if Git implemented precise merging it wouldn't be exponential time.
19:08 Riastradh Exponential time isn't the only thing.  `Why do you care?  Git works and it doesn't require a degree in physics to understand the groupoid structure of patches and their commutation.  It's an organized way to use diff & patch, and it works.'
19:10 jeltsch joined #darcs
19:10 copumpkin the whole "I don't have a degree in X, and don't want to need one to do Y" argument is annoying
19:10 roconnor Riastradh: :)  I think most people understand that they want merges to at least be associative.
19:11 roconnor most git users even
19:11 Riastradh I think most Git users don't care, roconnor.
19:11 roconnor they don't care that much
19:11 roconnor but given the choice I think they'd go for the associative version.
19:11 Riastradh `If you're worried about that, you're not paying enough attention to what the merge is doing.'
19:12 owst \whois Riastradh
19:12 owst well, damn
19:12 roconnor I still think most git users won't put forward that argument.
19:13 Riastradh Hi, owst.  I'm just some guy who uses and hates Git, and likes but doesn't much use Darcs.
19:14 roconnor I wonder if merging associatively isn't faster than 3-way merge since you don't have to do any fuzzy offsetting.
19:14 Riastradh I started using Darcs back in 2003 or so before 1.0 and before Git existed.
19:14 owst Heh Riastradh :). You seem to have some good arguments both ways, just wondering if there was a deep reason for that!
19:14 owst Yeah, that would be it.
19:17 roconnor Riastradh: you use linux?
19:17 Riastradh Not much, roconnor.  Only on other people's computers.
19:18 owst Usability question, if a user runs `darcs wh -l` and they have a malformed boring entry, should darcs a) continue, printing a warning (which may be lost, due to large `wh -l` output) or b) fail, notifying the user of the problem?
19:18 Riastradh Why can't the warning be printed at the end, owst?
19:19 owst Yeah, that just came into my head - I have a feeling it would be 'tricky' to keep the information around till after the tree has been diffed.
19:19 * owst peeks at that idea.
19:25 owst Yeah, doesn't look like it'd be neatly possible. So, fail early, or just continue, with a possibly hidden warning.
19:25 Riastradh How about `make it be neatly possible'?
19:30 owst I don't think I can... :( sucky.
19:47 lelit left #darcs
19:54 kowey_ joined #darcs
19:55 raichoo joined #darcs
19:56 kowey left #darcs
19:56 kowey_ is now known as kowey
20:07 gbeshers joined #darcs
20:12 owst Anyone got any ideas why the following loop would segfault when run through the test harness, but be fine when run manually:
20:12 owst for f in `seq 1 200`; do
20:12 owst echo $f > $f
20:12 owst darcs rec -lam $f
20:12 owst done
20:13 owst Wait, scratch that - it's my path.
20:15 owst Hmm, ignore the "scratch that" it still happens when I actually use the correct darcs.
20:18 Heffalump do you know what segfaults?
20:18 owst | test: line 13: 10186 Segmentation fault      darcs rec -lam $f
20:19 owst That's all I have.
20:19 Heffalump also, does it still do it with darcs rec --ignore-times -lam $f ?
20:19 * owst tries that.
20:20 owst Hmm, it goes much slower, and they are all skipped...
20:20 owst (but no failures)
20:32 Heffalump all skipped?
20:33 owst Sorry, I only ran the ones that failed.
20:34 Heffalump I'm still confused
20:34 owst Me too; let me explain again :)
20:36 owst Running all tests, get-http-packed.sh fails with a seg fault. I added --ignore-times and did `darcs test get-http-packed` then there are 4 tests that run, but they are all skipped. If however, I re-run the whole test suite it appears that the segfaults are still there.
20:37 owst Gah, sorry, the "skipped" is a red-herring, they do that anyway, if I just run `darcs test get-http-packed`, without adding --ignore-times.
20:38 abuiles joined #darcs
20:40 Heffalump biab, small child being difficult about bed...
20:41 owst :)
20:47 owst Heffalump: I have to disappear for a while, here's the change I've made to my repo, which apparently causes the seg fault: http://hpaste.org/45896. I just ran all tests on a plain get of screened and it skipped (but didn't segfault) on the get-http-packed.sh test.
20:55 arjanb left #darcs
21:12 lispy left #darcs
21:13 Heffalump all I can think is some kind of problem in the regex library.
21:14 Heffalump try enabling core dumps then using gdb to get a stack trace - see if it's coming from any library under Text.Regex
22:03 jeltsch left #darcs
22:11 raichoo left #darcs
22:21 iago joined #darcs
22:36 kowey left #darcs
22:44 abuiles left #darcs
22:52 copumpkin left #darcs
23:37 lispy joined #darcs
23:37 owst Well, debugging this is a pain.
23:38 owst Running via `cabal test` I get segfaults, each time - it does say core dumped, but I can't find the corefile, even though ulimit reports unlimited size for coredumps. If I run `valgrind cabal test` the tests are all Skipped, and no coredump
23:38 owst s/and no coredump/and no segfault/
23:49 iago owst, perhaps it is better to take a break
23:49 iago :P
23:49 owst Yeah, not getting far at the moment ;)
23:49 owst Doesn't help that I don't know much about debugging! :)
23:50 iago but which is the application that is giving you the segfaults
23:51 owst Well, I'm not sure - I think it's darcs
23:51 iago running the tests??
23:53 owst Sometimes! Sometimes, the tests just say "Skipped" which seems to be caused by the shell scripts doing `exit 200`, but even if I change the tests to not do that, it still happens!
23:55 owst WTF! using test -v causes the segfault again.
23:55 owst Man, this is weird.
23:55 iago but what are you doing in addition
23:55 iago I mean, I have never found a segfault running darcs tests
23:55 iago moreover, "cabal test" is not the way I run darcs tests
23:56 iago what are you doing concretely?
23:56 owst Oh, isn't it? How do you run them?
23:56 owst Well, I've added a patch to change the boring regex parsing.
23:56 iago you should use the dist/build/darcs-test/darcs-test binary
23:59 iago to be honest I don't know if "cabal test" calls to that binary
23:59 owst I think it might do, in some way.

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary